A fusion technique for iris localization and detection

An automated algorithm to localize irises for Middle East individuals had been developed in this research. Histogram equalization, Logabout, Difference of Gaussian (DoG), wavelet transformation, Principle Component Analysis (PCA) and Artificial Neural Network are popular techniques used for image processing, feature extraction and classification. A fusion of these techniques had been introduced to compensate the effects of illumination and head orientation for iris detection. The algorithm was tested with Middle East face database through experiments. In this paper, iris candidates are extracted from the valley of the detected face region after being pre-processed. All the detected iris candidates will go through wavelet transform. The wavelet coefficients are then reduced and extracted by PCA. Finally, Softmax Backpropagation Neural Network (SBNN) works as the iris classifier. The impact of the pre-processing techniques on the performance of the proposed algorithm was studied. The proposed algorithm had achieved a success rate of 90.5% with 0% false positives being reported.